Largely as a result of recent DNA exonerations, many of which had contained false confessions in evidence, a spotlight of scrutiny has been cast on the processes of police interviewing and interrogation. In recent years, a number of researchers have sought to systematically study various aspects of these processes (Kassin, 1997, Kassin & Gudjonsson, 2004; Gudjonsson, 2003; Leo, 1996a; Ofshe & Leo, 1997). Now, however, a steady stream of false confession stories emanating from New York, Los Angeles, Chicago, Miami, and other major U.S. cities are inundating newspapers and television news media on a regular basis (see Drizin & Leo, 2004; Leo & Ofshe, 1998). Many of these stories recount horrific tales of psychologically – and, in some cases, physically – abusive interrogations of children and adults, including many who were cognitively impaired. While it is necessary to shed light on the problem, as research has done, it is also important to know the extent to which the cases that come to public light represent practices that are “common” or “extraordinary.”

Forty years ago, the United States Supreme Court lamented the relative absence of empirical information about what constituted common or routine police interrogation practices. In its landmark decision in Miranda v. Arizona (1966), the Court wrote that, “Interrogation still takes place in privacy. Privacy results in secrecy and this in turn results in a gap in our knowledge as to what in fact goes on in the interrogation room” (p. 448). To fill this gap in its own analysis, this Court surveyed and critiqued then-existing police interrogation training manuals—most notably, Inbau and Reid’s (1962) Criminal Interrogation and Confessions, which is now in its fourth edition (Inbau, Reid, Buckley, & Jayne, 2001). Although the number of interrogation manuals and programs has increased over the last forty years (Leo, 2004), one cannot determine from them what constitutes common police practices.

Since Miranda, researchers have used an array of methods to fill in this empirical gap. This literature is comprised of individual case studies (e.g., Gudjonsson & MacKeith, 1990), archival analyses of actual case documents (e.g., Leo & Ofshe, 1998; Drizin & Leo, 2004), observations of taped interrogations (e.g., Moston, Stephenson, & Williamson, 1992; Ofshe & Leo, 1997), retrospective self-reports of suspects (e.g., Gudjonsson & Sigurdsson, 1999; Gudjonsson, 2006), and laboratory and field experiments (e.g., Kassin & Kiechel, 1996; Kassin, Meissner, & Norwick, 2005; Russano, Meissner, Narchet, & Kassin, 2005). In addition, a number of social scientists, journalists, and legal scholars have directly observed police practices “inside the interrogation room” (e.g., Corwin, 2003; Leo, 1996a; Simon, 1991; Skolnick, 1966; Wald Ayres, Hess, Schantz & Whitebread, 1967). As a result of this activity, there is now a substantial literature on various aspects of the psychology of police interrogations and confessions (Kassin & Gudjonsson, 2004; Gudjonsson, 2003).

A number of issues have attracted attention in recent research. One concerns the extent to which law enforcement professionals can accurately distinguish between truthful and deceptive statements. Police investigators have confidence in their ability to make such judgments during a pre-interrogation interview, judgments that often determine whether they interrogate suspects or send them home. This confidence stems from some combination of on-the-job experience and law enforcement training programs that promise to increase judgment accuracy. For example, John E. Reid and Associates claims that trained investigators can achieve an 85% level of accuracy through the use of various verbal cues (e.g., qualified or rehearsed responses), nonverbal cues (e.g., gaze aversion, frozen posture, slouching) and behavioral attitudes (e.g., lack of concern, anxiousness, and guardedness) presumably diagnostic of truth or deception. Indeed, field research confirms that police are confident in their skills, with some believing that they possess a “sixth” sense (Leo, 1996c).

The bulk of research indicates that this faith is misplaced. Over the years, numerous studies have demonstrated that individuals perform no better than chance at detecting deception, that training tends to produce only small and inconsistent increments in performance, and that “experts” perform only slightly better than ordinary people, if at all (for reviews, see Granhag & Strömwall, 2004; Memon, Vrij, & Bull, 2003; Vrij, 2000). Comparing the extent to which trained and naïve college students can distinguish between true and false denials to a mock crime, Kassin and Fong (1999) found that whereas exposure to the Reid technique failed to improve judgment accuracy, it significantly boosted confidence—a pattern also found among experienced investigators who were later tested using the same tapes (Meissner & Kassin, 2002). Apart from their performance in these experimental tasks, it remains to be seen whether investigators in general perceive themselves to be accurate judges of truth and deception.

A second issue concerns Miranda rights. In 1966, the U.S. Supreme Court ruled that prior to commencing interrogation police must inform custodial suspects of their Constitutional rights to silence and to appointed counsel, and that anything they say can be used against them. This Court further held that suspects must voluntarily, knowingly, and intelligently waive those rights in order for their statements to be admissible at trial. Subsequent rulings have carved out exceptions to this rule and limited the consequences for non-compliance, leading some scholars to question whether police are free to disregard Miranda (Clymer, 2002; Weisselberg, 1998; White, 2001; Zalman, 2003, 2004). One of the best known legal opinions in American history, Miranda is still a mainstay in the interrogation room. In two recent decisions, the U.S. Supreme Court has reaffirmed the basic warning-and-waiver requirement (Dickerson v. United States, 2000; Missouri v. Seibert, 2004), at one point proclaiming that Miranda is “embedded in routine police practice to the point where the warnings have become part of our national culture” (Dickerson v. United States, 2000, p. 2335).

In practice, the requirement may have little effect (Thomas & Leo, 2002). Over the years, police have adapted a number of strategies to circumvent or nullify Miranda and its invocation rules (Leo & White, 1999). For example, they may avoid the need for warnings by questioning suspects in settings that appear noncustodial (Leo, 1992). Or they may read the rights but then proceed to question suspects as though they had no choice in the matter, eliciting what some courts have called an “implicit waiver” (Leo, 2001a). Sometimes police properly read the Miranda rights and invocation rules but then downplay their significance and persuade suspects that it is in their own self interest to waive these rights (Leo, 1996b; Simon, 1991). In some cases, they interrogate “outside Miranda” by tricking suspects into talking after they had invoked their rights—producing “off the record” disclosures that may be used to generate other admissible evidence and to impeach the defendant at trial if he or she chooses to testify (Leo & White, 1999; Weisselberg, 1998).

Consistently, research has revealed that custodial suspects waive their rights approximately 80% of the time (Cassell & Hayman, 1996; Leo, 1996b; Wald et al., 1967). This high waiver rate stems not only from use of the aforementioned techniques but from the fact that many suspects—such as juveniles, people who are mentally retarded, and those who are under stress—do not fully understand their Miranda rights and how to apply them (Cloud, Shepherd, Barkoff, & Shur, 2002; Fulero & Everington, 2004; Grisso, 1981, Oberlander & Goldstein, 2001). Even among suspects who are competent, it appears that those who are innocent may be particularly disposed to waive their precious rights out of a naïve belief in the transparency of their innocence (Kassin, 2005; Kassin & Norwick, 2004).

A third issue concerns the interrogation tactics that are used to elicit confessions. The stated objective of interrogation is to move a presumed guilty suspect from denial to admission. The techniques used are thus designed to overcome a suspect’s resistance and to induce him or her to confess. The empirical study of police interrogation often begins with an examination of what the popular training manuals recommend (Leo, 1996a; Zimbardo, 1971). In addition, a number of researchers have studied the process directly, via firsthand observations and analyses of interrogation transcripts. These studies have shown that police use many of the recommended techniques—such as physical isolation, positive confrontations with evidence, and minimization. Putting this into a broader framework, Ofshe and Leo (1997) have suggested that interrogation can best be understood as a two-step psychological process in which the interrogator first seeks to convince the suspect that he or she is trapped and then attempts to induce that suspect to perceive that the benefits of confessing outweigh the costs.

A fourth issue concerns the amount of time suspects spend in the interrogation room, a factor of relevance to evaluating the voluntariness of the statements that are made. Observational studies have suggested that routine interrogations tend to be relatively brief encounters, with the modal duration ranging from 20 minutes to an hour (Cassell & Hayman, 1996; Leo, 1996a; Wald et al., 1967). These findings stand in stark contrast to Drizin and Leo’s (2004) archival study of 125 proven false confessions, where 34% of interrogations lasted 6 to 12 hours, 39% lasted 12 to 24 hours, and the mean was 16.3 hours. This latter finding is not surprising in light of prominent case studies of false confessions, so many of which were taken at night and after lengthy interrogations (e.g., Gudjonsson, 2003; Leo & Ofshe, 1998; Johnson, 1997; Ofshe, 1989). As observational studies are limited only to those investigations that researchers are invited or permitted to monitor, and as proven false confession stories represent an inherently biased sample of cases, it remains to be seen how experienced police investigators would characterize the norm when it comes to time-related issues.

A fifth issue concerns the rate at which suspects make incriminating statements to police in the form of admissions and full narrative confessions. Confession rates have always factored prominently into debates concerning law enforcement practices in the United States and elsewhere. Researchers have sought to determine confession rates to gauge how successful interrogators are (Gudjonsson, 2003; Leo, 1996a; Moston et al., 1992; Wald et al., 1967). Interrogation outcomes also represent an obvious “bottom line” for detectives who are largely evaluated by their ability to solve and close cases (Simon, 1991; Skolnick, 1966). The rate at which police elicit incriminating statements is also highly consequential for individual suspects and the criminal justice system as a whole. People who confess are treated differently at every subsequent stage of the criminal process than those who do not. They are more likely to be charged by prosecutors; more likely to be charged with more crimes; less likely to have their cases dismissed; more likely to have their cases resolved by plea bargaining; more likely to be found guilty, especially of more serious charges; and more likely to receive severe punishment upon conviction (Leo & Ofshe, 1998; Drizin & Leo, 2004).

In the United States, research has suggested that the confession rate ranges from 46% to 68% (Neubauer, 1974; Witt, 1973). Based on their own observations, Cassell and Hayman (1996) reported a confession rate of 54% in Utah; Leo (1996a) reported a confession rate of 64% in Northern California (for a review, see Thomas, 1996). Again, because these data are not derived from representative samples of interrogations or cases, it would be instructive to compare these rates to estimates that police investigators provide on the basis of their own collective experiences.

On the question of interrogation outcomes,  false confessions are a particular source of concern, arising with disturbing frequency in DNA exoneration cases. Although no one knows the extent of the problem or has devised an adequate method of calculating precise incidence rates, it is clear that the cases that have been discovered and publicized represent the tip of a much larger iceberg (Drizin & Leo, 2004; Gross, Jacoby, Matheson, Montgomery & Patel, 2005). Historically, the law enforcement community has been reluctant to recognize that people would confess to crimes they did not commit unless physically tortured or mentally disabled (Leo, 2001b). Is it possible that the recent spate of DNA exonerations and news media coverage of high-profile false confessions has softened police skepticism? Those who train interrogators have come to acknowledge the possibility in recent editions of their manuals (e.g., Inbau et al., 2001; Walters, 2003; Wicklander & Zulawski, 2002). What is not clear, however, is the extent to which practitioners share this concern.

A sixth issue concerns the ways in which police interrogations and confessions are recorded. Investigators are trained to convert the oral admissions they elicit into full written, audiotaped, or videotaped confessions that can later be presented in court. As more and more wrongful convictions based on false confessions are discovered, however, there has emerged a lively debate in state legislatures and in the courts about these record-keeping practices. Increasingly, there are calls for mandating the electronic recording of entire interrogations—a requirement that would arguably deter egregious interrogation tactics, reduce the number of false confessions, inhibit frivolous defense claims of coercion, and increase the fact-finding accuracy of judges and juries (for an overview of the history, issues, arguments, and current practices, see Drizin & Reich, 2004). In the past, most police departments in the United States resisted the recording of interrogations, arguing that the camera would inhibit suspects from talking, hinder police from using effective tactics that would offend judges and juries, and allow defense attorneys to attack the voluntariness of the confessions (e.g., see Inbau et al., 2001). It is interesting that police who have videotaped interrogations as a matter of policy or on a voluntary basis have reacted favorably to the practice (Geller, 1992; Buckley & Jayne, 2005; Sullivan, 2004). Still, it is not clear if there remains in general a culture of skepticism and resistance within the trenches of the law enforcement community.

The present survey

Although researchers have learned a great deal about interviewing, interrogation, and the elicitation of confessions, to date there has been no published survey of police interrogators themselves. In light of the steady stream of false confession stories in the news—some of which feature horrific tales of lengthy interrogations and the use of high pressure tactics—it is important to gain insight into police perspectives on how common the problem is, as well as the methods they use most often. To help fill this gap in our knowledge we designed a survey, the purpose of which was to describe the law enforcement perspective.

To address the specific issues and controversies previously described, we asked participants to estimate, rate, and otherwise self-report on six aspects of their work: (1) their ability to detect truth and deception; (2) Miranda warnings and waivers; (3) the use of various interrogation techniques; (4) the frequency, length, and timing of interviews and interrogations; (5) the rates of true and false confessions; and (6) their own practices and opinions with regard to the recording of interrogations and confessions. Across the board, our goal was to obtain reports directly from police investigators themselves, to distinguish, albeit imperfectly, common practices and beliefs from those that may be uncommon if not extraordinary, and to triangulate what has been found in other types of empirical research.

Method

Participants

The participants in this study were 631 investigators from 16 police departments in five American states (N = 574) and customs officials from two Canadian provinces (N = 57). The breakdown of participants by department is presented in Table 1. All participants were recruited from major jurisdictions on a voluntary basis, through a known contact, often a division commander, police chief, captain, or sergeant. They were then surveyed in individual or group settings arranged and sometimes administered by one of the authors. They were informed that the questionnaire would take about fifteen minutes to complete and were assured that they would not be asked for identifying personal information. All contacts were promised a copy of the results and assured that our goal was to determine what beliefs and practices are common—not to compare and contrast specific individuals or departments, or point fingers at those whose activities depart from the norm.

Table 1 Distribution of police departments represented in the sample
Full size table

The questionnaire

At the 2000 Meeting of the American Psychology-Law Society, introduced a survey instrument that he had developed and sought collaborators to help administer the survey to police in North America. The purpose was to obtain self-report data from law enforcement professionals with respect to their suspect interviewing and interrogation practices and beliefs on a range of important issues. Participants were not asked for identifying information other than sex, age, and state of residence; whether they were employed at the local, state or federal level; and whether they had received any special training (seminars, workshops, etc.) on how to conduct interviews and interrogations.

Following the demographic questions, participants were asked experienced-based questions concerning the percentages of all, guilty, and innocent crime suspects who (1) give a partial admission, (2) give a full confession, or (3) admit to nothing. In a later question, they also estimated they number of times they saw an innocent person confess (not only to police, but to friends or others) to a crime he or she did not commit.

In a number of questions, participants estimated the average number of times a suspect is interviewed or interrogated (from 0 to more than 10), the average length of an interrogation (in hours and minutes), the longest interrogation in which they were involved (in hours and minutes), and the percentage of interrogations distributed over various times intervals of day and night (8 a.m.-noon, noon-4 p.m., 4 p.m.-8 p.m., 8 p.m.-12 a.m., 12 a.m.-4 a.m., 4 a.m.-8 a.m.).

On the policy issue of recording interrogations, participants indicated whether their agency has a recording requirement (yes, no) and, if so, the nature of that requirement (videotaping, audiotaping, stenographic). Then they estimated the percentage of interrogations in which they were personally involved that were videotaped, audiotaped, or stenographically recorded, after which they indicated their own opinion as to whether and how interrogations should be recorded.

Turning to Miranda, participants were asked, in three separate questions, to estimate the percentage of people in general, guilty suspects, and innocent suspects who (1) fully waive their rights and submit to a complete interrogation, (2) initially waive but then invoke their rights at a later time, or (3) refuse to waive their rights from start to finish (for each question, participants were directed to provide all three estimates, totaling 100%).

With regard to their own performance and behavior, participants were asked to estimate how accurate they are at knowing when a suspect is telling the truth or lying (% correct, incorrect). With specific regard to taking confessions, they estimated the percentage of times they took an open narrative or question-and-answer statement that was written by the investigator and signed by the suspect; both handwritten and signed by the suspect; audiotaped; and videotaped. Finally, participants were asked to estimate how often they had employed various interrogation techniques that are recommended and/or used in law enforcement (1 = never, 2 = on rare occasion, 3 = sometimes, 4 = often, 5 = always). These techniques, and accompanying results, are listed in Table 2.Footnote 1

Table 2 Self-reported frequency of usage of 16 techniques on a 1 (never) to 5 (always) scale
Full size table

Results

Results of the survey were screened for evidence of normality in the distribution of responses (significant Skewness or Kurtosis). If non-normal distributions were found, outliers were identified and omitted from the analysis. Both uncorrected and corrected estimates are reported for interested readers. Mean, median, range, and standard deviation estimates are also reported below where appropriate. Some participants neglected to respond to one or more items; as a result of these missing data, sample sizes differed somewhat across questions and are reported accordingly.

Six hundred thirty one investigators filled out the survey: 531 males (86%) and 88 females (14%). Overall, 571 (91%) of the participants were employed in local police departments and 54 (9%) were in federal law enforcement agencies (two worked for the state; one indicated “other”). On average, participants were 40.68 years old (Med = 40; Range = 21 to 62; SD = 7.83; N = 616), had worked for an average of 16.37 years in law enforcement (Med = 16; Range = 1 to 38; SD = 7.67; N = 624), and had conducted—by their own estimates—a mean of 1,675 suspect interviews and interrogations over the course of their careers (Med = 500; Range = 2 to 100,000; SD = 5,589; N = 584). This latter estimate of interrogations conducted demonstrated a significant degree of Skewness (11.8, SE = .10) and Kurtosis (179.9, SE = .20). To correct for normality of the distribution, significant outliers were determined by creating a fence based upon 3 interquartile ranges. Omission of outliers successfully normalized the distribution and suggested an average of 723 suspect interviews over the course of their careers (Med = 500; Range = 2 to 3,000; SD = 751; N = 530), representing an average of 46.3 interviews conducted per year. As might be expected, those responses identified as outliers involved investigators with significantly more years of experience when compared with the normalized sample (t = 5.02, p < .001; Ms = 20.98 vs. 15.62; SDs = 6.60 vs. 7.57; Ns = 54 vs. 530, respectively).

It is generally assumed that all law enforcement investigators receive special training in interviewing and interrogation techniques, typically when they are promoted from patrol officer to detective. The Chicago-based firm of John E. Reid & Associates states that it has trained more than 300,000 law enforcement professionals since 1974 and that it continues to train thousands more every year (see http://www.reid.com; last accessed on May 5, 2006). In our sample, 517 (82%) participants said they had received special training (seminars, workshops, etc.) on how to conduct interviews and interrogations. Of those who had received special training, 11% reported that their training was in the Reid technique. This finding seems surprisingly low since it has been widely assumed by both the courts (Miranda v. Arizona, 1966) and the research community (Zimbardo, 1971; Leo, 1996a; Kassin & Gudjonsson, 2004) that the Reid technique is the most prevalent method of police interrogation training in the United States. It is important to note, however, that trained respondents may not have recalled the specific type of training they had received. Hence, our results may underestimate the number of those who were Reid-trained.

Truth and lie detection

To determine how confident investigators are in their lie detection abilities, we asked participants to estimate the percentage of times that their personal judgments of suspects’ veracity turn out to be correct. According to Vrij (2000), accuracy rates among professionals obtained in research laboratories have ranged from 45% to 60%, with a mean of 54%. Suggesting overconfidence, in contrast, our participants on average estimated that they can distinguish truthful and deceptive suspects at a 77% level of accuracy (Med = 80; Range = 0 to 100; SD = 16.65; N = 607).

Miranda requirements

Participants were asked to indicate the medium by which they most often apprised suspects of the Miranda rights and elicited a waiver of those rights. The results suggested that the most common methods were to inform suspects orally (67%) or in writing (29%), but seldom on audiotape (3%) or videotape (1%). With regard to the Miranda decisions that suspects make, our investigators estimated that 81% of “people in general” waive their Miranda rights (Med = 85; Range = 0 to 100; SD = 16.69; N = 612). Of the 81% believed to waive their Miranda rights, participants estimated that 68% submitted to complete interrogations (Med = 70; Range = 0 to 100; SD = 21.24; N = 612), whereas 13% initially waived their rights but then invoked (Med = 10; Range = 0 to 85; SD = 13.12; N = 612).

Participants were also asked to separately estimate, from their own experience, how both guilty and innocent suspects react to Miranda. Consistent with the hypothesis that innocent suspects are more likely to forego existing safeguards, participants perceived a difference—estimating on average that 73% of guilty suspects waive their rights (Med = 80; Range = 0 to 100; SD = 22.88; N = 589) compared to 84% of innocent suspects (Med = 90; Range = 0 to 100; SD = 22.33; N = 540).

Interrogation practices

To assess the variety of interrogation tactics that police use, we asked participants to estimate on a 1–5 point scale, the frequency with which they have used each of 16 techniques in trying to get suspects to confess. In descending order of frequency, Table 2 lists these techniques and their self-reported frequencies of usage.

The portrait that emerges from these self-reports is that the typical interrogation consists almost always of isolating the suspect away from family and friends (62% reported they “always” use this technique); placing him or her in a small private room (“always” = 41%); identifying contradictions in the suspect’s account (“always” = 40%); and trying to establish rapport with the suspect in order to gain his or her trust (“always” = 31%). The typical interrogation often, but not always, includes confronting the suspect with evidence of his or her guilt and appealing to his or her self-interests. Somewhat less frequently, interrogation sometimes includes offering the suspect sympathy, moral justifications and excuses; interrupting the suspect’s denials and objections; implying and/or pretending to have independent evidence of guilt; minimizing the moral seriousness of the offense; and appealing to the suspect’s religion or personal conscience. On rare occasions interrogators show the suspect photographs of the crime scene and/or victim (28% reported they “never” used this technique); express impatience, frustration or anger at the suspect (“never” = 29%); threaten the suspect with consequences for not cooperating (“never” = 49%); and have the suspect take a polygraph exam and tell him that he failed it (“never” = 48%). Last on the list, participants indicated that they almost never physically intimidated the suspects whom they interrogated (“never” = 71%).

A factor analysis was conducted on these 16 ratings to determine whether certain techniques cluster together, revealing patterns of techniques that are frequently used in conjunction with one another or are conceptually related. An exploratory factor analysis with Varimax rotation was conducted to assess whether orthogonal factors might be found within the data.  Four factors with eigenvalues >1.0 were identified, accounting for 53% of the data.  Fourteen of the items were found to load >.50 onto one of the four factors.  Cross-loadings for items in each factor are presented in Table 3. The first factor accounted for 19% of the variance and included tactics related to “Isolation, Rapport, and Minimization” (sympathy, self interest, religion, rapport, isolation, and minimization). The mean frequency of usage of this factor was 3.52 (Med = 3.67; Range = 1 to 5; SD = 0.68). The second factor accounted for 13% of the variance and included elements of “Confrontation” (identifying contradictions, confrontation, and interruptions). The mean frequency of usage of this factor was 3.79 (Med = 3.67; Range = 1 to 5; SD = 0.66). The third factor accounted for 11% of the variance and involved “Threatening the Suspect” (threats of punishment, frustration and anger, and physical intimidation). The mean frequency of usage of this factor was 1.78 (Med = 1.67; Range = 1 to 5; SD = .70). The fourth and final factor accounted for 10% of the variance and consisted of two items relating to the “Presentation of Evidence” (false polygraph feedback and crime scene photographs). The mean frequency of usage of this factor was 2.08 (Med = 2.00; Range = 1 to 5; SD = 0.90). All factors were statistically distinguished in their rate of usage (all pairwise ts (614) \(\geq\!\)7.57, ps < .001): Elements of “Confrontation” were employed significantly more often than all other factor groupings, followed by “Isolation, Rapport, and Minimization,” “Presentation of Evidence,” and “Threatening the Suspect.”

Table 3 Cross-loadings of interrogation items in rotated factor solution
Full size table

Several regression models were generated to assess whether individual differences in respondents’ characteristics predicted their self-reported use of each set of techniques. Five predictor variables were included in the models: (a) the number of years the respondent had served in law enforcement (experience), (b) whether or not the respondent had attended special training in interview and interrogation (training), (c) the respondent’s confidence in his or her own lie detection ability (confidence), (d) the number of interrogations conducted by the respondent (interrogations), and (e) the average length of interrogations conducted by the respondent (length).

Results of the regression models (see Table 4) confirmed previous research on the role of experience, training, confidence, and length of interrogations on deception detection ability (e.g., see Meissner & Kassin, 2002). In particular, respondents with greater experience in law enforcement were more likely to advocate the use of “Threatening the Suspect” (r = .09) and “Presentation of Evidence” tactics (r = .19). Those who had attended special training in interrogation were significantly more likely to employ both “Isolation, Rapport, and Minimization” tactics (r = .14) and the “Presentation of Evidence” (r = .15). Respondents who expressed greater confidence in their deception detection ability were more likely to employ “Isolation, Rapport, and Minimization” tactics (r = .18) and “Confrontation” (r = .19). Those who had conducted a greater number of interrogations were significantly more likely to advocate “Threatening the Suspect” (r = .09), but were less likely to employ “Confrontation” tactics (r = −.12). Finally, respondents who generally conducted longer interrogations were more likely to support the use of “Isolation, Rapport, and Minimization” tactics (r = .13) and the “Presentation of Evidence” (r = .11).

Table 4 Results of regression models predicting interrogation usage from experience, training, deception detection confidence, and number and length of interrogations conducted
Full size table

Length, frequency, and timing of interrogations

As observational studies are limited to the sessions that researchers are permitted to monitor, and as false confession stories represent an inherently biased sample of cases, it is important to know what police consider an average length of interrogation—and what they recall as their longest interrogation. We also wanted to know how often they tend to question suspects and during what times of day. On the first question, participants estimated, based on their own experience, that the mean length of interrogation is 1.60 hours (Med = 1; Range = 0 to 7; SD = .89; N = 601). In response to the second question, they estimated that their longest interrogation had lasted an average of 4.95 hours (Med = 4; Range = 1 to 72; SD = 5.72; N = 606). This latter estimate demonstrated significant deviation from normality in estimates of Skewness (7.02; SE = .10) and Kurtosis (68.80; SE = .20). Omission of outliers successfully normalized the distribution and suggested an average of 4.21 hours for the longest interrogations conducted by respondents (Med = 4; Range = 1 to 13; SD = 2.47; N = 588).

Individual suspects are often questioned more than once in a single investigation and their statements may result from a process that extends over some period of time. Hence, we asked participants to estimate from their own experience the average number of times a suspect is interviewed or interrogated (0–10) as well as the time of day of these sessions (in six 4-hour intervals). To the first question, participants estimated that suspects are questioned 3.08 times per investigation (Med = 2; Range = 1 to 10; SD = 2.23; N = 590). As to time of day, they estimated that roughly 82.88% of interrogations were conducted during typical waking hours (Med = 90; Range = 0 to 100; SD = 20.69; N = 568). In particular, respondents estimated that 20.08% were conducted from 8 a.m. to noon, 20.35% from noon to 4 p.m., 21.77% from 4 p.m. to 8 p.m., and 22.48% from 8 p.m. to midnight. In contrast, an estimated 17.14% of interrogations were held during typical sleep hours (Med = 10; Range = 0 to 100; SD = 20.99; N = 563), including an estimated 11.67% from midnight to 4 a.m. and 5.67% from 4 am to 8 a.m.

Confession rates

Turning from the process of interrogation to its outcomes, we asked participants to estimate from personal experience the percentage of their suspects who gave a partial admission, full confession, or no self-incriminating statement. Though estimates were somewhat high relative to prior observational studies, the results closely paralleled these studies. On average, participants estimated that an average of 67.57% of suspects had made self-incriminating statements (Med = 75; Range = 0 to 100; SD = 22.23; N = 621), with 38.40% providing partial admissions and 30.01% providing full confessions. An estimated 32.14% of suspects make no admissions or confessions (Med = 25; Range = 0 to 100; SD = 32.14; N = 621).

We then asked participants to separately estimate confession rates for suspects who were guilty and innocent. Predictably, these questions drew radically different estimates. Respondents estimated that 69.48% of guilty suspects provide a confession (Med = 75; Range = 0 to 100; SD = 23.91; N = 584). Separately, they estimated that 40.24% provided partial admissions and 33.83% provided full confessions. In contrast, they estimated that 23.30% of innocent suspects provide some form of confession (Med = 5; Range = 0 to 100; SD = 33.33; N = 524). A significant degree of Skewness (2.30; SE = .11) and Kurtosis (4.24; SE = .22) was noted in the distribution of false confession estimates. Omission of outliers succeeded in normalizing the distribution and suggested that an average of 4.78% of innocent respondents provide a false confession (Med = 0; Range = 0 to 30; SD = 7.66; N = 380; this estimate was significantly greater than 0, z = 12.16 in a one-sample z test), with an average of 3.80% of innocent suspects providing a partial admission and 0.97% providing a complete false confession.

We sought similar information in a more direct subsequent question that asked participants to indicate the number of times they had seen an innocent person confess—whether to investigators, friends, or others. To this question, participants reported having seen an average of 2.97 false confessions (Med = 0; Range = 0 to 500; SD = 21.49; N = 593). A significant degree of Skewness (21.20; SE = .10) and Kurtosis (486.37; SE = .20) was noted in the distribution of estimates. Omission of outliers succeeded in normalizing the distribution and suggested an average of 0.71 false confessions observed (Med = 0; Range = 0 to 6; SD = 1.36; N = 543; this estimate was significantly greater than 0, z = 12.17 in a one-sample z test).

Recording of interrogations and confessions

Participants were asked about the format they use for taking confessions (written by the investigator, handwritten by the suspect, audiotaped, or videotaped), the practices of their agencies, and their opinions on the matter. With regard to the narrative confessions, there was a great deal of variability in reported practices. Overall, participants estimated that 56.25% of the confessions they have taken were in written form (Med = 70; Range = 0 to 100; SD = 41.14; N = 551), with 33.96% of statements written for the suspect to sign and 23.67% personally handwritten by the suspect. In contrast, respondents estimated that 37.92% of confessions were electronically recorded (Med = 20; Range = 0 to 100; SD = 39.54; N = 553), with 31.41% audiotaped and 8.19% videotaped.

Regarding full interrogations, we first asked participants to indicate whether their agencies required that suspect interviews and interrogations be recorded: 16% said yes, 84% said no. Of those police departments with a recording requirement, the most common method was audiotape (59%), followed by videotape (25%) and stenographic recording (16%). When asked about the interrogation sessions in which they had been involved, participants estimated that 8.51% were fully videotaped, 35.82% were audiotaped, 14.49% were transcribed by stenographer, and 42.38% were not recorded in any way. Finally, we asked participants about their opinions on whether interviews and interrogations should be fully recorded from start to finish and 81% said yes. Within this group, 51% favored the videotaping of interrogations, 42% favored audiotaping, and 7% favored a written transcript.

Discussion

Our survey of police investigators, the first of its kind, was designed to get law enforcement perspectives about various aspects of suspect interviews, interrogations, and confessions. Several interesting findings were obtained—some that were consistent with past research, others that were completely new.

On the subject of deception detection, we asked investigators about their accuracy at judging the veracity of suspects—a pivotal judgment, regularly made, that can determine whether they proceed to interrogate suspects or send them home. Over the years, studies have shown that law enforcement professionals, like laypeople, are poor discriminators of truth and deception and that training does not reliably improve performance (Granhag & Strömwall, 2004; Memon et al., 2003).

Following prior reviews of the empirical literature, Bond and DePaulo (2006) estimated that people exhibit on average and across studies only a 54% accuracy rate (47% for lies, 61% for truths)—a baseline that challenges the uncorroborated claim that training in the use of verbal and nonverbal deception cues can increase accuracy rates to 85% (Inbau et al., 2001). In recent years, experiments have shown that police are significantly more confident but not more accurate than lay people at distinguishing true and false denials (Meissner & Kassin, 2002) or confessions (Kassin et al., 2005). When we asked participants in this study to rate their own deception detection skills, they estimated a 77% level of accuracy, a figure that substantially exceeds human lie detection performance and parallels these earlier findings.

Ever since the U.S. Supreme Court imposed the Miranda warning and waiver requirements, questions have been raised about the practical effects, or lack of effects, on police and suspects (see White, 2001). Research suggests that roughly four out of five people waive their rights. In our survey, participants’ self-reported experiences were highly consistent with this finding, as they estimated an overall waiver rate of 81%–a figure that closely tracks naturalistic data, representing the exact midpoint between the waiver rate found in Leo’s (1996b) observational study in Northern California (78%) and Cassell and Hayman’s (1996) study of interrogations through prosecutorial screening sessions in Utah (84%). Interestingly, our respondents estimated that 68% waived their rights fully, whereas 13% initially waived their rights but later invoked. This estimated high rate of initial waiver followed by invocation had not previously been observed (only 1% in Leo’s sample; 4% in Cassell & Hayman’s sample). Our result thus suggests that police perceive Miranda as more of a safeguard during interrogation (i.e., after initially waived) than is actually the case. In their estimates, respondents also exhibited the belief that innocent suspects are more likely than offenders to waive their rights (84% to 74%). This perceived disparity corroborates Kassin and Norwick’s (2004) laboratory-based finding and the resulting argument that Miranda may not fully protect those citizens who need it the most (Kassin, 2005).

When participants self-reported on their use of various interrogation techniques, the results closely paralleled Wald et al.’s (1967) early observational study in which “the most common approach was to confront the suspect with evidence or with the assertion that there was a witness… The confrontation might be accompanied with the admonition that the detectives had all the information they needed to convict the suspect, and that he would make it easier for all concerned if he would fill in the rest of the story” (p. 1543). The present results are also generally consistent with Leo’s (1996a) observation that interrogators tend to use the techniques found in training manuals and seldom resort to tactics that the courts have deemed coercive—most notably, threats and the use of physical intimidation.

With regard to the infrequent use of high-pressure tactics, two important points are worth noting. First, this result is based on self report, so although it is consistent with some observations (Leo, 1996a; Uviller, 1988), it indicates less coercion than has been observed by others (Simon, 1991; Wald et al., 1967). The second point is that participants said they sometimes use two tactics that, although legally accepted, can lead innocent people to confess: pretending to have independent evidence of guilt and minimization strategies (Ofshe & Leo, 1997). In laboratory studies, false confession rates to non-crimes are increased both by the false evidence ploy (e.g., Horselenberg, Merckelbach, & Josephs, 2003; Kassin & Kiechel, 1996; Redlich & Goodman, 2003) and by comments designed to minimize the apparent seriousness of the crime or culpability of the suspect (Russano et al., 2005). As these techniques have been sanctioned by both the U.S. and Canadian Supreme Courts (Frazier v. Cupp, 1969; R. v. Oickle, 2000) and are recommended in popular training manuals (e.g., Inbau et al., 2001), it is disappointing but not shocking that they are reportedly used with some degree of regularity.Footnote 2

We also examined the conceptual clustering of techniques used by respondents in everyday practice. Results of an exploratory factor analysis demonstrated that techniques involving “Confrontation” were employed most frequently followed, respectively, by “Isolation, Rapport, and Minimization,” “Presentation of Evidence,” and “Threatening the Suspect.” We also conducted several regression models to assess the predictive nature of investigator background characteristics. Consistent with previous research indicating that prior training and experience are associated with a propensity to make judgments of deception and guilt (Meissner & Kassin, 2002, 2004), these models indicated that more years of law enforcement experience, special training, greater confidence in one’s deception detection ability, and a history of conducting more and longer interrogations all predicted greater endorsement of one or more sets of interrogation techniques. Thus it would appear that investigator characteristics associated with a tendency to presume guilt are predictive of more frequent use of psychologically manipulative and confrontational techniques in everyday practice.

Related to the use of various tactics, we asked participants about temporal aspects of interrogation. Based on their own experience, they estimated that the mean length of interrogation is 1.60 hours, which is consistent with the amount of time deemed necessary by those who train interrogators (Inbau et al., 2001) and somewhat higher than the average reported in observational studies (Cassell & Hayman, 1996; Leo, 1996a; Wald et al., 1967). Investigators also estimated that their longest interrogation lasted an average of 4.21 hours—which is far shorter than the 16.3 hour average found in the population of false confession cases studied by Drizin and Leo (2004). Taken together, these normative data provide a critical anchor point for evaluating individual cases involving confessions taken after many hours of interrogation—as seen in the investigations of the Central Park jogger defendants, the Phoenix Temple Four defendants, Michael Crowe, Joshua Treadway, Tom Sawyer, Gary Gauger, Corey Beale and other exonerated false confessors. Quite clearly, the lengthy sessions in these cases were extraordinary, not normative or common practice.

The time of day or night during which a suspect is interrogated, not just the length of interrogation, is also relevant to an analysis of voluntariness and coercion. Overall, participants estimated that roughly 83% of interrogations were distributed across typical waking hours, from 8 a.m. to midnight, with only 17% being conducted during typical sleep hours, from midnight to 8 a.m. As so few interrogations extend over the post-midnight hours, and as research shows that fatigue and sleep deprivation heighten susceptibility to influence and impair complex decision-making (Blagrove, 1996; Harrison & Horne, 2000), it is reassuring to see that this practice—seen in several high-profile false confessions—is a non-normative exception to the rule.

When it comes to the outcomes of interrogation, participants estimated that 68% of suspects had made self-incriminating statements, a number that is high but consistent with prior observational studies with confession rates ranging from 46% to 68% (in our survey, the estimate included both partial admissions and full confessions). This result indicates that detectives see themselves as generally quite successful at eliciting admissions and confessions. Regardless of how accurate these estimates are, they suggest that law enforcement officers of the twenty-first century do not feel “handcuffed,” as their predecessors had once feared, by the Miranda warning and waiver requirements. Indeed, this result parallels two generations of empirical “Miranda impact” studies suggesting that Miranda does not hinder police in their efforts to interrogate suspects and elicit confessions (Thomas & Leo, 2002).

Of particular interest were the results obtained when we asked participants to separately estimate the self-incrimination rates for guilty and innocent suspects. As one would reasonably expect, participants reported a far higher rate among suspects who were guilty (69%). The reported self-incrimination rate among innocent suspects also exceeded zero by a substantial margin, at 24%. However, omission of outliers to normalize the distribution lowered the mean false confession rate to 4.78%, which was significantly greater than 0. Within this group, 3.80% provided a partial admission and only 0.97% provided a full confession. We also asked respondents to more directly indicate the number of times they had seen an innocent person confess—to investigators, friends or others. To this question, they reported an average lifetime total of 2.97 false confessions, down to 0.71 when outliers were omitted—an estimate that was also significantly greater than 0.

On the important policy question of whether police investigators themselves favor or oppose the videotaping of interrogations, our results clearly corroborate the recent strong law enforcement support for electronic recording that has been reported elsewhere. Indeed, although only 16% of the participants in our survey worked in jurisdictions where the electronic recording of interrogations was required (none were in states in which it was mandatory), 81% believed that interviews and interrogations should be fully recorded, from start to finish. This finding echoes and extends the strong police support for electronic recording first reported by Geller (1992) and more recently found both in states that require videotaping (Buckley & Jayne, 2005) and in departments that have chosen to do so voluntarily (Sullivan, 2004).Footnote 3 In these other studies, the favorable reports were from investigators who were themselves recording interrogations. In the present survey, favorable attitudes were also reported by those not required or experienced at recording interrogations. Hence, this finding appears to reflect a generalized and growing sentiment within the law enforcement community that electronic recording is no longer a policy to be feared but one to be embraced—“a policy whose time has come” (Donovan & Rhodes, 2000; Kassin & Gudjonsson, 2004).

Although the present research breaks new ground as the first published survey of police investigators in North America, there are methodological shortcomings that may limit the generalizability of our findings. First, although a large number of participants were tested, we did not randomly select them from the broad universe of police investigators in the United States and Canada. Instead we employed convenience samples, at least in some cases, testing investigators from 16 police agencies with whom the authors were able to establish contact and secure cooperation. Ultimately a national random survey that draws from all states should be conducted. Second, we asked participants in several questions to estimate the frequencies of various practices (e.g., the percentages of interviews and interrogations they had conducted, how often they used particular techniques, how often they recorded interrogations) and outcomes (e.g., the percentage of suspects who waived or invoked their Miranda rights, or who made self-incriminating statements). Research shows that except for behaviors that are rare and important, people are unlikely to have detailed representations of single instances in memory, leaving them to rely on inference strategies that can reduce the accuracy of their estimates (Schwarz, Groves, & Schuman, 1998; Tourangeau, Rips & Rasinski, 2000). In studies that assess accuracy, correlations between survey estimates and objective records data have ranged from .25 to .86 (Burton & Blair, 1991).

Despite these potential limitations, it is worth noting that on survey questions for which potentially corroborating data are available, responses were highly consistent with results elsewhere obtained. Asked about the percentage of their suspects who waived their Miranda rights, participants provided a mean estimate of 81%, which is precisely at the midpoint of what others had found (Cassell & Hayman, 1996; Leo, 1996). Consistent with recent laboratory research, they also estimated the waiver rate to be higher among innocent suspects than among those who turned out to be guilty (see Kassin & Norwick, 2004). Similarly, participants estimated from their experience a 68% self-incrimination rate—which closely resembles statistics derived from observational studies (see Thomas, 1996). Considering the social desirability biases that can skew self-reports, one may speculate that participants under-reported some estimates (e.g., the frequency with which they threaten or physically intimidate suspects; the number of false confessions they have witnessed) and over-reported others (e.g., how accurate they are at distinguishing truth and deception; how often they elicit confessions from guilty suspects). Importantly, however, our participants were anonymous and explicitly assured that their responses to the self-administered questionnaire were confidential—data collection methods that tend to minimize social desirability biases (DeMaio, 1984; Tourangeau et al., 2000).

While the present survey provides new information about investigators’ self-reported interrogation practices, beliefs, and perceptions, our knowledge of contemporary police interrogation remains incomplete. Over the last forty years, researchers have made great strides in addressing the “gap problem” that the U.S. Supreme Court inveighed against in Miranda v. Arizona (1966). The Miranda Court noted that this problem stems from secrecy in the interrogation room. As electronic recording becomes increasingly commonplace, however, there should be less and less secrecy about the process. In an era of electronic recording, the ideal way for scholars to measure and study actual police practices and their outcomes—and thus the best way to find out what is common practice and what is extraordinary—is to observe large numbers of videotaped interrogations, randomly selected from across the country, involving a full range of crimes. While we hope that this article will inspire researchers to conduct surveys to assess police investigators’ perceptions and beliefs, we also hope that police agencies that record their interrogations will make their tapes available for systematic research on interviewing and interrogation practices and their outcomes.